Willis Eschenbach at Watts Up With That discovers homogenization shenanigans in the Global Historical Climate Network temperature record for Darwin, Australia, excerpted below:

Intrigued by the curious shape of the average of the homogenized Darwin records, I then went to see how they had homogenized each of the individual station records. What made up that strange average shown in Fig. 7? I started at zero with the earliest record. Here is Station Zero at Darwin, showing the raw and the homogenized versions.
What on earth justifies that adjustment? How can they do that? We have five different records covering Darwin from 1941 on. They all agree almost exactly. Why adjust them at all? They’ve just added a huge artificial totally imaginary trend to the last half of the raw data! Now it looks like the IPCC diagram in Figure 1, all right . . . but a six degree per century trend? And in the shape of a regular stepped pyramid climbing to heaven? What’s up with that?
Those, dear friends, are the clumsy fingerprints of someone messing with the data Egyptian style . . . they are indisputable evidence that the “homogenized” data has been changed to fit someone’s preconceptions about whether the earth is warming.
One thing is clear from this. People who say that “Climategate was only about scientists behaving badly, but the data is OK” are wrong. At least one part of the data is bad, too. The Smoking Gun for that statement is at Darwin Zero. . . .
Now, I want to be clear here. The blatantly bogus GHCN adjustment for this one station does NOT mean that the earth is not warming. It also does NOT mean that the three records (CRU, GISS, and GHCN) are generally wrong either. This may be an isolated incident, we don’t know. But every time the data gets revised and homogenized, the trends keep increasing. . . .
And with the Latin saying “Falsus in unum, falsus in [omnis]” (false in one, false in all) as our guide, until all of the station “adjustments” are examined, adjustments of CRU, GHCN, and GISS alike, we can’t trust anyone using homogenized numbers.